Someone asked me last week why their post ranks third but Google still answers the question above their link. That question is the whole game now. Featured snippets don't reward the best page. They reward the clearest answer to a specific kind of question. And most people are still writing pages when they should be writing answers.
I've been tracking this since snippets first started stealing clicks, and the shift I keep coming back to is this: the keyword you chase and the keyword Google needs are rarely the same shape. One is a phrase. The other is a question with a format attached. If you're still building your content around noun phrases, you're leaving the top of the page to someone else.
Key Takeaways
- Question-based keywords win snippets because the query itself tells you the format Google wants.
- Definition questions ("what is") want a short paragraph. Process questions ("how to") want steps. Comparisons want a table.
- You find these questions in autocomplete, the "questions" filters in search tools, and your own Search Console query report — not by guessing.
- Snippets and AI Overviews are now two doors into the same answer. You can be eligible for both with one well-structured paragraph.
- Schema markup helps, but the plain-text answer on the page is what actually gets lifted.
- You don't need to rank first to earn the snippet. You need to answer best.
What makes question-based keywords different from ordinary ones
A normal keyword tells you the topic. A question-based keyword tells you the topic and the shape of the answer expected. That second part is what everyone ignores, and it's exactly where the advantage sits.
Think about the difference between "email deliverability" and "why do my emails go to spam." The first is a subject. The second is a request with a built-in format: a cause, explained plainly, in a few sentences. If your page opens with a 300-word history of SMTP, you've already lost the slot.
The format signal hidden in the query
Google has been matching question type to answer format for years now, and once you see the pattern you can't unsee it. Here's the matrix I actually use when mapping content:
| Question type | Example | Format that gets pulled |
|---|---|---|
| Definition | What is a featured snippet? | Short paragraph, 40–50 words |
| Process | How to optimise for featured snippets? | Numbered or bulleted list |
| Comparison | Snippet vs AI Overview | Table or two-column list |
| Quantitative | How many words for a snippet? | A single figure, bolded |
| Causal | Why did my rankings drop? | Short paragraph, cause first |
Definition and causal questions almost always resolve into a paragraph. Process questions resolve into a list. The moment you know which bucket your target question falls into, your formatting decision is basically made for you.
What most writers do wrong is pick one format for the whole article and force every question into it. That's backwards. Let the question dictate the container, not the other way around.
What are some good SEO keywords?
Good SEO keywords are the ones that match real search intent and carry enough specificity to be winnable. For snippets specifically, the best ones are questions people actually type, phrased the way they type them, with a clear answer that fits in a compact block.
In practice, a "good" keyword for this purpose has four traits. It's long enough to signal intent rather than a broad topic. It maps to one of the format types above. It's answerable in under 60 words without losing meaning. And it brings some commercial or informational value — otherwise you're optimising for traffic that bounces.
Where to actually find these questions
I'll be blunt: guessing at questions is the most common waste of time I see. There are four places that reliably surface real phrasing, and none of them require paid tools to start.
- Autocomplete in the search bar. Type your topic, then a space, then "why", "how", "what", "can". The suggestions that complete are real query strings.
- Search tools that offer a dedicated questions filter — these surface related interrogatives directly.
- Your own Search Console query report, filtered to strings starting with a question word. This is the most underused asset you have.
- Your support inbox and comment sections. If a customer asked it in an email, someone typed a version of it into a search bar.
That last point deserves more weight than it gets. Two years ago I started logging every question from customer emails into a single spreadsheet. Within a few months it was beating every keyword tool I'd tried, because the phrasing came from real people with real problems, not from an estimate.
What are the good keywords if you're starting from scratch?
Build a list of 15 to 20 questions on your topic first, then test them. Don't optimise yet. Assign each one a format based on the table above, and mark which you could answer better than the current top result. The ones you can answer better are your shortlist. Everything else is noise.
Are featured snippets still relevant?
Yes, but the way they work has changed, and anyone telling you they're dead is reading the wrong signal. Featured snippets still appear above organic results and still capture the first answer a user reads. What's shifted is the competition for that real estate.
Since AI Overviews rolled out, a single question can now be answered in two places on the same page: the overview, and the snippet. In my experience, the content that gets cited in an overview tends to be the same content structured for a snippet. Clear, direct, well-scoped answers. The tactics overlap far more than people assume.
Snippet vs AI overview — what's the practical difference?
The snippet lifts your answer almost verbatim. The overview synthesises across sources, so you might be cited or you might not, even if you're eligible. That asymmetry matters: a snippet is a near-guaranteed placement, while an overview is a competition for a mention.
Which brings up an obvious problem. If the overview answers the question before the user scrolls, your snippet's click value drops. The numbers I've seen on my own pages bear this out — some question keywords that used to send steady traffic now drive far fewer clicks, even when the placement is held. The answer is visible. The click isn't needed.
Does that mean you stop? No. It means you stop chasing snippet placements that only produce visibility and start prioritising questions where the user still needs to go deeper. Comparison queries and process queries still pull clicks, because the reader wants the detail an overview won't give them. Pure definition queries are the ones bleeding.
What is the 80/20 rule in SEO?
The 80/20 rule in SEO is the observation that a small share of your pages tends to generate most of your organic traffic, and a small share of your keywords tends to drive most of your clicks. It's a heuristic, not a law — but it holds often enough to plan around.
Applied to snippets, it means you shouldn't spread effort evenly across every question on your topic. A handful of well-chosen questions will do the heavy lifting. My own pattern is consistent: I put real effort into maybe five questions per article cluster. The rest get a line or two and move on.
How to apply it without being lazy
- List every question you could plausibly answer on the topic.
- Score each on search volume signal and how well it matches your actual expertise.
- Keep the top fifth. Format those properly, with schema and a clean answer block.
- Treat the rest as supporting text — useful for context, not for chasing placement.
The trap is treating 80/20 as permission to do less work. It isn't. It's permission to concentrate the same effort in fewer places, which is usually harder than spreading it thin.
How to optimise for featured snippets?
You optimise for featured snippets by answering the target question directly, in the format the question implies, near the top of the page. Everything else is secondary.
Here's the order that works, based on what I've tested:
- Put the answer immediately after the heading. Not after a story, not after context. The first paragraph under the H2 is the one that gets lifted.
- Keep paragraph answers to roughly 40–50 words. Long enough to be complete, short enough to be quoted.
- Use the question itself as your H2 or H3, phrased the way people search it.
- Format process answers as a real ordered list. Tables for anything comparative.
- Add FAQ or HowTo structured data where it genuinely applies. It helps eligibility, though it won't rescue a weak answer.
- Bold the single figure or key term that carries the answer.
One thing I'd flag from experience: the answer block does not need to be the first thing on the page. It needs to be the first thing under the relevant heading. I've seen pages win snippets with the answer sitting 800 words deep, as long as the heading matched and the paragraph was clean.
The mistake I made for months
I spent a stretch rewriting articles to cram an answer block right at the top, near the introduction. Placement rate barely moved. What actually shifted things was renaming the heading to match the query phrasing exactly and tightening the paragraph below it. The location was never the problem. The match was.
Do you need schema markup to win a snippet?
No, and it's worth being honest about that. Schema can improve your odds and clarify intent to the crawler, but the plain text answer is what gets extracted and displayed. I've earned snippets on pages with no structured data at all, simply because the paragraph was clean and the heading matched. Schema is a helper, not a key.
Where this leaves you
The uncomfortable part of all this is that question-based keywords force you to write more precisely than most content teams are used to. No warm-up. No "in this section we'll look at." Just the answer, in the shape the question asked for, placed where the crawler expects to find it.
Which raises a question worth sitting with: if a clear 45-word answer is what earns the placement, how much of what you published this year was actually built to be the answer — and how much was built to be read? Those aren't the same thing anymore. The pages that win the top slot are the ones written for the first reader Google sends, not the human who might arrive later.
